1. Field of the Invention
The present invention relates to a reproduction signal evaluation method and reproduction signal evaluation unit using a PRML signal processing system, and an optical disk device adopting the same.
2. Description of the Background Art
Recently the shortest mark length of recording marks have reached the limit for optical resolution, and an increase in inter-symbol interference and deterioration of SNR (Signal Noise Ratio) are becoming obvious as the density of optical disk media increases, therefore the use of a PRML (Partial Response Maximum Likelihood) system as a signal processing method is becoming common.
The PRML system is a technology combining partial response (PR) and maximum likelihood (ML) decoding, and is a known system for selecting a most likely signal sequence from a reproduced waveform, assuming the occurrence of inter-symbol interference. As a result, it is known that decoding performance improves compared with a conventional level decision system. (e.g. see Blue-ray Disk Books, supervised by Hiroshi Ogawa and Shinichi Tanaka, Ohmsha Ltd, Dec. 10, 2006)
On the other hand, a shift in signal processing systems from a level decision to PRML has resulted in generating some problems in reproduction signal evaluation methods. Jitter, that is a reproduction signal evaluation index, which has been used conventionally, is based on the assumption that signals are processed using a level decision system. This means that in some cases jitter has no correlation with the decoding performance of the PRML system, of which signal processing algorithms are different from the level decision system. Therefore a new index having correlation with the decoding performance of the PRML system has been proposed (e.g. see Japanese Patent Application Laid-Open No. 2003-141823 and Japanese Patent Application Laid-Open No. 2004-213862).
A new index to position shift (edge shift) between a mark and a space, which is critical for recording quality of an optical disk, has also been proposed (e.g. see Japanese Patent Application Laid-Open No. 2004-335079). If a PRML system is used, this index must also be correlated with the performance of the PRML, and must quantitatively express the shift direction and quantity of the edge for each pattern, according to the concept of the PRML system.
As the density of magnetic disk media increases further, the problem of inter-symbol interference and SNR deterioration becomes more serious. In this case, the system margin can be maintained by using a higher level PRML system (e.g. see Blue-ray Disk Books, supervised by Hiroshi Ogawa and Shinichi Tanaka, Ohmsha Ltd, Dec. 10, 2006). In the case of an optical disk medium of which diameter is 12 cm and recording capacity per recording layer is 25 GB. The system margin can be maintained by using a PR1221 mL system, but in the case of a 33.3 GB recording capacity per recording layer, a PR12221 ML system must be used. In this way, it is expected that the tendency to use a higher level PRML system would continue in proportion to the increase in densities of optical disk media.
Japanese Patent Application Laid-Open No. 2003-141823 and No. 2004-213862 disclose using “a differential metric, which is a difference of the reproduction signals between the most likely first state transition sequence and the second most likely second state transition sequence” as the index value.
If there are a plurality of patterns of “a most likely first state transition sequence and second most likely second state transition sequence” which have the possibility of causing an error, these patterns must be statistically processed systematically. This processing method is not disclosed in Japanese Patent Laid-Open No. 2003-141823 and No. 2004-213862. Japanese Patent Application Laid-Open No. 2003-272304 discloses a method for detecting a plurality of patterns of “a differential metric of reproduction signals between the most likely first state transition sequence and the second most likely second state transition sequence” detected in the same manner as in Japanese Patent Application Laid-Open No. 2003-141823 and No. 2004-213862, and the processing of a pattern group.
In PR12221 ML signal processing, which is disclosed in Japanese Patent Application Laid-Open No. 2003-272304, there are three types of patterns which easily cause an error (pattern group of merging paths of which Euclidian distance is relatively short). In this pattern group. The pattern generation probability and the number of errors, when the pattern generates errors occur in a pattern, differ depending on the pattern, so according to Japanese Patent Application Laid-Open No. 2003-272304, a standard deviation a is determined from the distribution of the index values, which are acquired for each pattern, and the errors to be generated are predicted based on the generation probability of the pattern (generation frequency with respect to all parameters) and the number of errors to be generated when the pattern has an error.
In Japanese Patent Application Laid-Open No. 2003-272304, a method for assuming the distribution of the acquired index values as a normal distribution and predicting a probability for the index value becoming “0” or less based on the standard deviation σ thereof and variance average value μ, that is, a probability of generation of a bit error, is used as an error prediction method. This, however, is a general method for predicting error generation probability. The method for calculating the predicted error rate according to Japanese Patent Application Laid-Open No. 2003-272304 is characterized in that generation probability is determined for each pattern, the predicted error rate is calculated, and this predicted error rate is used as a guideline of signal quality.
However, with the method according to Japanese Patent Application Laid-Open No. 2003-272304, the error rate cannot be predicted accurately if recording distortion occurs to recording signals. This problem becomes particularly conspicuous when data is recorded by thermal recording, such as the case of an optical disk, since recording distortion tends to be generated by thermal interference. As the density of optical disk increases, space between recording pits decreases even more, and an increase in thermal interference is expected, therefore this problem will be unavoidable in the future. The problem of the predicted error rate calculation method according to Japanese Patent Application Laid-Open No. 2003-272304, which cannot appropriately evaluate the signal quality of signals having recording distortion, will now be described.
FIG. 21 shows an example of frequency distribution of a differential metric of a specific pattern, which is used as a signal index in Japanese Patent Application Laid-Open No. 2003-141823 and No. 2003-272304. Generally speaking, the spread of the distribution of the differential metric is caused by the noise generated in an optical disk. The reproduction noise generated in an optical disk is random, so this distribution usually is a normal distribution. And this differential metric is defined as a “differential metric of the most likely first state transition sequence and second most likely second state transition sequence”, and is a distribution of which center is a square of the Euclidean distance between the most likely first state transition sequence and the second most likely second state transition sequence of an ideal signal (hereafter defined as the signal processing threshold). The standard deviation of which center is this signal processing threshold is the index value defined in Japanese Patent Application Laid-Open No. 2003-141823, No. 2004-213862 and No. 2003-272304. The probability of this differential metric becoming 0 or less corresponds to the predicted error rate. This predicted error rate can be determined using the inverse function of the cumulative distribution function of this normal distribution.
FIG. 21A is a distribution diagram when no substantial distortion occurred during recording, and FIG. 21B and FIG. 21C show distribution diagrams in a state where recording edges in the recording pits shifted due to thermal interference during recording, and recording distortion occurred. If distortion occurs due to thermal interference, the frequency distribution of the differential metric of a specific pattern becomes a normal distribution of which center value is shifted. This shift of the center position corresponds to the distortion generated by thermal interference. FIG. 21B and FIG. 21C are cases when a predetermined amount of shift occurred in the plus or minus direction from the center of the distribution, and an index value to be determined is the same value for both FIG. 21B and FIG. 21C, and the index value increases since the center of the distribution has shifted. An increase in the index value should mean an increase in the probability of error generation, but errors decrease in the case of FIG. 21C.
This is because in the case of FIG. 21B, where the center of the distribution is shifted to the side closer to “0”, error generation probability (probability of differential metric becoming 0 or less) increases, but in the case of FIG. 21C, where the center of the distribution is shifted to the plus side, error generation probability decreases. This reversal phenomena is because an error is generated only when the index value based on the differential metric approaches 0, which is the major difference from the jitter of the time axis, that is the index value conventionally used for optical disks. In the case of a conventional jitter of the time axis, errors increase regardless the side, plus or minus, to which the center position of the distribution shifts, therefore the above mentioned problem does not occur.
A problem similar to the above also occurs in the case shown in FIG. 21D. FIG. 21D is a case when the determined distribution of the differential metric is not normal distribution. This occurs when the thermal interference during recording is high, and thermal interference is also received from the recording marks before and after “the most likely first state transition sequence and second most likely second state transition sequence”. The thermal interference amount is different depending on the length of the recording marks before and after, and the shift of recording mark positions generates a differential metric distribution where two normal distributions (distribution 1 and distribution 2) overlap.
In distribution 2, where there is a shift to the plus side from the signal processing threshold, error generation probability drops, but the index value, which is a standard deviation from the signal processing threshold as the center, increases because of the influence of distribution 2. Just like the case of FIG. 21C, error rate also decreases when the index value increases. In this way, if the prior art reported in Japanese Patent Application Laid-Open No. 2003-141823 and No. 2003-272304 is applied to a high recording density optical disk of which thermal interference is high, the correlation of the index value and error rate worsens.
An idea for solving this problem is disclosed in Japanese Patent Application Laid-Open No. 2003-51163. This is a method of counting a number of differential metrics with which the differential metric, acquired from a predetermined pattern group, which becomes smaller than a predetermined threshold (e.g. half of signal processing threshold). A method for determining a predicted error rate based on this count value is also disclosed. In the case of this method, a side closer to 0 of the differential metric distribution, that is the side which has a possibility of generating an error, is used for the evaluation target, so the above mentioned problems in Japanese Patent Application Laid-Open No. 2003-141823 and No. 2003-272304 do not occur. But a new problem, mentioned herein below, occurs, since a predetermined threshold is used and a number of differential metrics exceeding this threshold is measured. This problem will be described with reference to FIG. 21E.
FIG. 21E shows an example of counting the differential metrics of the distribution which exceeds the threshold, which is half of the signal processing threshold. The differential metrics less than this threshold are counted, and the ratio of the parameter of pattern generation and the count value is used as the signal index. If it is assumed that the distribution of the differential metric is a normal distribution based on this count value, the probability when the differential metric becomes smaller than 0 can be determined, and the predicted error rate can be calculated. FIG. 21F shows an example of frequency distribution when the signal quality is good (signal quality with about an 8% jitter). In such a case, the spread of the distribution of the differential metric becomes narrow, and the number of differential metrics which exceeds the threshold decreases dramatically.
In the case of FIG. 21F, only about 0.2%, out of the differential metric distribution, can be measured. This means that a wide area must be measured in order to increase the accuracy of the measurement, which increases the measurement time and diminishes measurement stability. Also if there are defects and scratches generated during manufacture of the disks or if there is dust on a disk surface, the differential metric is generated in an area not greater than the threshold due to this defect (illustrated in FIG. 21F). In such a case, a number of the differential metrics, which exceed the threshold generated in the normal distribution, cannot be counted correctly. An advantage of conventional time axis jitter used for optical disks is that it is not affected by such defects, since standard deviation of measured time fluctuation is used and all the measured data is used.
The method disclosed in Japanese Patent Application Laid-Open No. 2003-51163, on the other hand, does not have this advantage of the conventional method based on time axis jitter, which is not affected by the defects, and therefore has a problem when used for the index values of optical disks where such defects as scratches and fingerprints easily occur. In order to increase the number of differential metrics to be measured using the method according to Japanese Patent Application Laid-Open No. 2003-51163, the threshold could be increased, but if the threshold is increased, another problem occurs, that is the accuracy of the predicted error rate drops. In an extreme case, if the threshold is increased to half of the Euclidean distance, a number of differential metrics that exceed the threshold becomes half of the number of measured samples, therefore it no longer depends on the spread of distribution, and accurate measurement becomes possible. In this way, in the case of the method according to Japanese Patent Application Laid-Open No. 2003-51163, the value of the threshold must be adjusted in order to maintain constant measurement accuracy depending on the quality of measured signals, and such adjustment is possible if the manner of how distribution spreads is somewhat understood, nonetheless this is a major problem for optical disks, where signal quality changes significantly.
Japanese Patent Application Laid-Open No. 2003-51163 and No. 2003-272304 also disclose a method of using bER predicted by the differential metric as the index, but if this is used as an index value, compatibility with the time axis jitter, which has been used as the signal quality evaluation index of optical disks, is lost, and handling is difficult.